Source. I've often read books that talk about a rise or a decline in learning during a given period, but this lecture was the first time I'd heard someone give such concrete numbers. What I'd really like is a book or article that analyses any period in the West (including, if possible the Sassanian and/or Islamic worlds) between say 200 and 1500 - the broader and more concrete the better.
What he is referring to are the specific manuscripts from the period that survives to this day, not the total sum of manuscripts that existed in that period.
Obviously there would have been many more manuscripts in existence during the period of 550-750, but the majority has since disappeared for the reasons that Freedman himself lists.